Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Tue, 25 Oct 2022 19:34:05 -0700
From:   Jakub Kicinski <kuba@kernel.org>
To:     Horatiu Vultur <horatiu.vultur@microchip.com>
Cc:     <netdev@vger.kernel.org>, <linux-kernel@vger.kernel.org>,
        <davem@davemloft.net>, <edumazet@google.com>, <pabeni@redhat.com>,
        <UNGLinuxDriver@microchip.com>
Subject: Re: [PATCH net 3/3] net: lan966x: Fix FDMA when MTU is changed
Message-ID: <20221025193405.1c8f6e74@kernel.org>
In-Reply-To: <20221023184838.4128061-4-horatiu.vultur@microchip.com>
References: <20221023184838.4128061-1-horatiu.vultur@microchip.com>
        <20221023184838.4128061-4-horatiu.vultur@microchip.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Precedence: bulk

On Sun, 23 Oct 2022 20:48:38 +0200 Horatiu Vultur wrote:
> +		mtu = lan_rd(lan966x, DEV_MAC_MAXLEN_CFG(port->chip_port));

I'm slightly confused about the vlan situation, does this return the
vlan hdr length when enabled? or vlans are always communicated
out-of-band so don't need to count here?

Unrelated potential issue I spotted:

        skb = build_skb(page_address(page), PAGE_SIZE << rx->page_order);       
        if (unlikely(!skb))                                                     
                goto unmap_page;                                                
                                                                                
        dma_unmap_single(lan966x->dev, (dma_addr_t)db->dataptr,                 
                         FDMA_DCB_STATUS_BLOCKL(db->status),                    
                         DMA_FROM_DEVICE);  

Are you sure it's legal to unmap with a different length than you
mapped? Seems questionable - you can unmap & ask the unmap to skip
the sync, then sync manually only the part that you care about - if you
want to avoid full sync.

What made me pause here is that you build_skb() and then unmap which is
also odd because normally (if unmap was passed full len) unmap could
wipe what build_skb() initialized.